HS 2.0-rc2A (win32) seems to apply associative rule like:
    (x[n]*y)*z --> x[n]*(y*z)
to optimize cycles like:
    for (int n=0; n<x.length; n++) {
        . . . x[n]*y*z . . .
    }

Note however, that in general case (x*y)*z is NOT equal to x*(y*z) in Java;
so that such optimization is illegal.

There are two tests in JCK 1.2.2 intended to reveal such illegal optimization:
    lang/FP/fpl036/fpl03605m3/fpl03605m3.html
    lang/FP/fpl036/fpl03605m3/fpl03605m4.html
(fpl03605m3 is for 'float', and fpl03605m4 is for 'double' arguments).

These tests fail against HS 2.0-rc2A (win32) when executed several times
in the same JVM.

To reproduce this bug, please try the scripts 'doit3.bat' and 'doit4.bat'
in the directory: /net/sqesvr/vsn/GammaBase/Bugs/<this_bug_ID>


Name: elC94547 Date: 03/16/2000



I have just observed the failure of these JCK 1.2.2 tests
against HS 2.0-rc2D (win32) installed over JDK 1.3.0-rc2W.

It looks like the failure is only reproduced when the tests
are stressed; in particular - when they are involved by the
stress-test:
    nsk/stress/jck122/jck122002
belonging to the testbase:
    /net/sqesvr/vsn/VM/testbase/testbase_nsk

To reproduce this failure, please try the script DOIT.BAT
at the directory:
    /net/sqesvr/vsn/GammaBase/Bugs/4310299


======================================================================